Nowadays it seems that artificial intelligence has to offer a solution for everything. Those who have not set any steps implementing some form of Artificial Intelligence might get the impression that they are doomed. Well, you don’t have anything to worry about. Supply chain environments have way more challenges ahead that must be tackled before AI really can help companies move forward.
Artificial Intelligence (AI) and machine learning are two buzzwords that everyone likes to use nowadays. But what do they actually mean? Many companies proudly claim to be working with AI, but in practice they do things that have been around for years and have little to do with new technology. We first need to have a good understanding of how AI exactly works.
From feedback loop to Deep Learning
AI consists out of three dimensions. The first level is relatively simple and we call it the ‘feedback loop’. We use data to teach something to a machine and allow it to work more efficiently based on this learning process. In many organizations this process is sufficient to speak of AI, but this is no longer so groundbreaking.
The second, deeper level includes machine learning. In this case, the machine learns from its own behavior and consequently gets better. A simple example is a computer that we learn to play tic-tac-toe. In the beginning you win easily, but the system remembers why it lost and it will never repeat that pattern again. After a while, the computer just can’t be beaten anymore.
The most far-reaching level is deep learning. This involves training a machine to think for itself by mimicking the neuronal networks of a human brain. For example, if you show a computer enough images of cats, the machine will eventually be able to recognize a cat in any picture.
Everything starts with data
Deep learning is still completely unrealistic for most companies. It requires a lot of data and it is not always certain whether you have access to all necessary information. Moreover, it is not enough to provide a machine with lots of data if you do not define the purpose very precisely.
For example, let’s say you launch a promotion and you want to estimate how much you’re going to sell. The promotion will be affected by the promotion set up, the weather, actions of competing brands, etc. So you need to have access to all these data before you can start to think about making correct predictions.
Only the bigger companies today have enough data to use AI in a good way. That is why companies must ask themselves what exactly they expect from AI. What do they want to achieve with it and will this technology move them forward?
AI is still in an experimental phase and most companies are far from ready to use this kind of technology. Of course companies need to think about how they can use AI to their advantage in the future. It is essential to collect the right data from now on.
Keep things under control
What we should definitely avoid is that AI becomes a black box. We secretly dream of creating a machine that is an extension of our own. However, we should never completely trust technology. We should always keep asking ourselves why it makes a certain decision. For example, a system can tell you when a machine will shut down, but do we know why this is the case?
Actually, we are also confronted with this kind of decisions in daily life. Let’s take a look at the Waze traffic application. It can advise you to take the next highway exit, many people will simply follow the suggested route. Others may be critical on Waze’s suggestion and based on their own knowledge, experience and the timeframe they will simply ignore Waze’s advice.
Either way, you arrive at work a little later at worst. But if a company makes the wrong decision based on data, the consequences can be much greater and the CEO will have to explain this to the shareholders. Therefore it is necessary to keep an overview on why certain decisions were made.
Focus on the essentials
Instead of investing heavily in AI, companies better focus on their business processes first. Research shows that 83 percent of executives say that a lack of visibility in their business processes is a bottleneck to apply AI in their supply chain. So what better way to start identifying these issues at hand?
To identify these obstacles, you don’t need AI at all. You can already start working way more efficiently with data on hand. Example: a local company believes that it can save costs by purchasing material in China, because it is much cheaper over there. After a simulation with data, it appears that it is better to pay a little bit more for that material in the Netherlands and to make more profit overall. Data on hand will easily tell you this but many companies still make similar mistakes, because they have way too little insight into their processes.
Companies must therefore first reach the necessary maturity before they can start working with AI. Many SMEs currently have too little time to tackle things thoroughly, let alone implement AI. Collecting data is a good start, but it is wiser to make step-by-step progress. Actually we are still crawling but we already have the ambition to run the 100m at the Olympic games. When companies are ready to grow, there is still plenty of time to look at AI’s real potential.